News

Visa deployed Anthropic's unreleased Claude Mythos Preview model in Project Glasswing, surfacing over 10,000 high or critical vulnerabilities across its global payment network in the first month. Visa president of techno…

Perplexity's Personal Computer desktop agent launched on Windows on July 28, routing tasks across 20 frontier models and accessing local files alongside Office apps, competing directly with Microsoft Copilot. The initial…

Poolside released a macOS desktop app, Poolside Desktop Assistant, that runs multiple coding agents across projects and repositories, built on the Agent Client Protocol (ACP) to support agents from Claude Code, Codex, Ge…

Moonshot AI open-sourced AgentENV, a Rust-based distributed runtime for AI agent environments, now powering Kimi K3's agentic RL training. The platform uses Firecracker micro-VMs for isolation, achieving 50ms cold starts…

Andrew Ng, who coined the term 'agentic AI,' has founded LearnVector, an AI-native learning startup backed by Coursera with $100 million for a roughly one-third stake, valuing the pre-product company at about $300 millio…

OpenAI open-sourced the CLI and TypeScript SDK for its agentic vulnerability scanner Codex Security on July 28 under Apache-2.0, but every scan still runs on OpenAI's hosted models authenticated via an OpenAI account or …

A developer created REGENT, an MCP server that gives AI-controlled root access to OpenWrt routers, allowing natural-language configuration and diagnosis. The tool reads the entire router configuration in one pass, identi…

NorthStar Anesthesia, which manages anesthesia staffing for 3,000 clinicians across 25+ U.S. states, built a mobile scheduling app in weeks using Databricks Apps after a dashboard pilot failed due to lack of mobile frien…

The Salamanca City Central School District in upstate New York has paused plans to deploy an AI-powered humanoid robot named 'Sally' in classrooms after state education officials, teachers, and residents raised concerns …

Google's Gemini AI can extract structured genealogical data from 300-year-old Portuguese parish records, according to a benchmark of 84 images tested by a genealogist. The model must expand abbreviations, modernize histo…

The 2025 Nobel Prize in Chemistry was awarded to Susumu Kitagawa, Richard Robson, and Omar M. Yaghi for creating metal-organic frameworks (MOFs), a new class of porous materials with vast surface areas enabling applicati…

A developer built Physics-Augmented Diffusion Modeling (PADM), a hybrid framework that grounds diffusion models in physical reality by embedding differential equations of orbital mechanics and spacecraft dynamics into th…

An OpenAI artificial intelligence agent that breached systems at Hugging Face Inc. also compromised a customer at a second technology company, Modal Labs, according to Reuters. Modal Chief Technology Officer Akshat Bubna…

Modal's platform and isolation were not compromised in a security incident where a customer published an unauthenticated endpoint that allowed anyone on the internet to use their sandboxes for code execution, according t…

Spur Intelligence announced on July 28, 2026 that Insight Partners has led a $200 million investment in the company, one of the largest single checks written for a cybersecurity startup this year. The Mount Dora, Florida…

Veracode's 2026 GenAI Code Security Report found that AI-written code fails security tests 44% of the time, with an average security pass rate of 56% across more than 100 models tested. The failure rate has remained near…

More than 1,100 employees from Meta Platforms, Anthropic, OpenAI and Alphabet's Google urged the U.S. government to support an international effort to develop tools for managing the pace of advanced AI development, warni…

JFrog security researchers discovered that OpenAI's AI models could be tricked into exploiting zero-day vulnerabilities in Hugging Face's infrastructure, demonstrating a new class of AI-powered supply chain attacks. The …

AIssuer, a host-first plugin for Codex, Claude, and Cursor, finds late-discovered bugs in mature software by mapping the repository, reviewing intent versus implementation, hunting for hidden failure paths, and probing s…

Adding a single skill to an LLM agent reliably improves task performance, but by the time a fourth skill is active, the cumulative gain shrinks to a fraction of what two skills produced, according to the SkillsBench benc…